User forums > General (but related to Code::Blocks)

GCC 4.4.1-tdm-2 for MinGW (with installer)

<< < (6/8) > >>

nenin:

--- Quote from: TDragon on October 30, 2009, 08:56:28 pm ---Have you observed similar problems in the official MinGW release(s)?

--- End quote ---
So, bad news. I reproduced problem on official MinGW 4.4.0, on 4.4.0-tdm-1-dw2 and on 4.4.1-tdm-2-dw2.
4.3.1-tdm-1-dw2  and 4.3.3-tdm-1-dw2 works properly. Usage of the "-O1"... "-O3" with  4.4.* leads to generation fallacious code  while the application of the full set of stand-alone  switches like "-fauto-inc-dec -fcprop-registers -fdce -fdefer-pop -fdse -fguess-branch-probability -fif-conversion2 -fif-conversion -finline-small-functions -fipa-pure-const -fipa-reference -fmerge-constants -fsplit-wide-types -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copyrename -finline-functions -funswitch-loops -fpredictive-commoning -fgcse-after-reload -ftree-vectorize -fthread-jumps -falign-functions -falign-jumps -falign-loops -falign-labels -fcaller-saves -fcrossjumping -fcse-follow-jumps -fcse-kip-blocks -fdelete-null-pointer-checks -fgcse -fgcse-lm -findirect-inlining -foptimize-sibling-calls -fpeephole2 -fregmove -freorder-blocks -freorder-functions -frerun-cse-after-loop -fsched-interblock -fsched-spec -fschedule-insns -fschedule-insns2 -fstrict-aliasing -fstrict-overflow -ftree-switch-conversion -ftree-pre -ftree-vrp -ftree-dce -ftree-dominator-opts -ftree-dse -ftree-fre -ftree-sra -ftree-ter -funit-at-a-time" does not.

Max:
nenin,

I observed that problem in several occasions. Using official MinGW and TDM as well. In most of the cases the prolem was not in compiler but in the source code. Writing floating point code is not trivial. Especially if you are playing with precions and using 386 processor family (80 bit FPU register). Try the options

-O2  -mfpmath=sse -msse2

if the problem disappears review your code. Anyway we are OT, not related to CB.

Max

nenin:

--- Quote from: MaxGaspa on November 10, 2009, 09:57:52 pm ---nenin,
<***>
if the problem disappears review your code.
Max
--- End quote ---
Could you please read more carefully what I wrote?  :roll:

jaxon:

--- Quote from: nenin on November 10, 2009, 07:38:22 pm ---
--- Quote from: TDragon on October 30, 2009, 08:56:28 pm ---Have you observed similar problems in the official MinGW release(s)?

--- End quote ---
So, bad news. I reproduced problem on official MinGW 4.4.0, on 4.4.0-tdm-1-dw2 and on 4.4.1-tdm-2-dw2.

--- End quote ---
It's very interesting what about Linux version of gcc...

Max:
Nenin wrote,,
>Could you please read more carefully what I wrote?

Yes I did. Unfortunately (please don't get me wrong) statements like

>...leads to generation fallacious code
>...I observe a lost of precision in numeric code if "O" flags are used.

really mean nothing :-)

You should post a small testcase to be analyzed. The only thing I get is that you have problems with "precision" (how much? which one? what?). Because using finite precision the results depend on the order of instructions (and optimizer is reordering/deleting instruction) I suggested (at least) to try using IEEE754.

Post, not here, a small testcase (for example try to report the observed issue in the bug archive in the TDM site).


Bye

Max

Navigation

[0] Message Index

[#] Next page

[*] Previous page